I am trying to develope a simple web-base front end tool which
analysis clickstream of a site. I am looking for an open api which can
discover and execute the cube.

Possible options include following:
1) OLE DB (MS)
2) JOLAP (Oracle, IBM, Hyperion)
3) XML/A (MS, Hyperion, SAS)

XML/A seems a good candidate although there might be performance
overhead. It doesn't tide to any language or platform. Since I only
knows Java, have anyone implement any client tool with java and apache
AXIES api? Anybody knows if Hyperion has release the implemetation for
XML/A?

I am not looking at JOLAP because ORALCE and IBM aren't big market
players. Also, I know for sure that hyperion doesn't have the
implemtation for the JOLAP althogh Hyperion claim to join the force of
implementing JOLAP. Maybe, hyperion doesn't have implementation for
XML/A as well? If that's the case, I have a guts feeling that I might
have to my tool without OLAP server due to the chaotic of the OLAP
api. But I have very little clued of implemeting a such thing. A star
chema is a good start but it's not quite yet enough. The dataware
house will contain mammoth of data which loaded data from web server
and back end server logs. Therefore, it must scale and performce well.
So, I expect that I might have to implement some very sophisticated
storage, any recommandation. I don't know what else I can said but I
think I only knows very little about OLAP, hope this can lead to a
good discussion. Thanks in advance.

There's an interesting statement: "I am not looking at JOLAP because
ORALCE and IBM aren't big market players."

Anyway, if you're interested in the Microsoft APIs but don't consider
XML/A mature, why don't you look into insulating your application from
the underlying API by going with a product like Simba's that add a layer
of abstraction.

Seth


liber wrote:

Another consideration might be, with Oracle OLAP at least, is to
consider the API that is called 'SQL'. It is after all probably the
most 'open' API you could think of.

You can work directly with multidimensional data, and the
multidimensional calculation engine of the OLAP Option in Oracle using
PL/SQL calls and simple SQL SELECTs etc.

Just a suggestion. Kevin (@ Oracle).

Seth Grimes wrote:

"Kevin Lancaster" <KevinDotLancaster (AT) Oracle (DOT) Com> wrote


I personally don't want to flatten my OLAP data into
2-d rowsets with SQL. What I want to have in my client application,
is a subcube (a multidimensional structure).

I think it's time for Oracle to stop spreading that "SQL access
to MD data is the best thing after, well, the last thing which was
really great" nonsense.

Hi there mar (AT) notexist (DOT) com

I'm sorry that you personally think that making multidimensional OLAP
data and calculations accessible via SQL is nonsense. That is your
choice. Liber will make his too - but I hope he will at least consider
it as one of the options he evaluates.

What I can report is that the Oracle customer-base certainly seem very
interested in this feature (which is one of the very good reasons for
having done it). As are vendors such as Business Objects who have
found that the SQL Access route opens up some very interesting
possibilities for their users.

Perhaps we can agree the following :

1) If you want a subcube (a multidimensional structure) in your client
application, that is entirely up to you. Making it easy for you to
populate that with 'rows' via SQL is a useful additional feature,
surely. You don't have to use it if you don't want to. PL/SQL calls
allow you to pull back the data as arrays if you prefer.

Most standalone MOLAP cubes I am aware of (eg Oracle Express, Hyperion
Essbase, Microsoft Analysis Services, IBM DB2 OLAP etc); and the leading
Query and Reporting tools which have deployed a middle or client 'cube'
tier; generally are loaded with data sourced from a relational data, and
therefore SQL is quite widely used already. If you can now also use it
to get data (stored and derived) from a multidimensional data source in
the same database, I would have thought that many people would find that
helpful.

Plus ofcourse there are thousands, millions probably, of people out
there that can use or develop in SQL.. rather more than the people that
know how to code to the proprietary APIs and query languages of Express
Server, Essbase/DB2OLAP and Microsoft Analysis Services put together I'd
guess. I happen to think that as more an more people figure this out,
it will greatly broaden the reach and applicability of OLAP.

2) The use of SQL as an API/access method does not make the underlying
data or calculations 'flat' - just the output. I suppose I just don't
get your point here. I wonder if you are confusing using SQL as an API
to fetch data and results of multidimensional calculations etc performed
on the server, with SQL that contains all the calculation logic in the
SQL Query using SQL syntax and applies it to relational tables.

3) Liber, in his orginal question said "I am trying to develope a simple
web-base front end tool which analysis clickstream of a site. I am
looking for an open api which can discover and execute the cube."

Sounds like an ideal fit with OracleDB and the OLAP option to me.
Clickstream data is the kind of detail data you would generally store
relationally, and build multidimensional cubes from it for analysis.
So, having both detail data and multidimensional summaries in the same
database, accesible by the same 'API' would seem a good fit - which is
why I made him aware of the SQL access to Oracle OLAP.

Hope it helps/clarifies. Kevin (@ Oracle)

mar wrote:

Seth Grimes <grimes (AT) altaplana (DOT) com> wrote

You said your looking for a tool. I assume you don't necessarily want
to build one from scratch. Why don't you take a look at TM1 WEB from
Applix. TM1 web provides an Excel interface to OLAP cubes over the
web. If you can use Excel and you have the data in a flat file or
relational database. You can have your application running over the
web in 2-3 hours.